Model
Model explorer

Qwen3-8B (Non-Thinking)

OPEN
Alibaba · Qwen3 family · released Apr 29, 2025

Fast non-thinking mode variant of the same 8B dense weights, for low-latency dialogue.

ReasoningCodingVisionFunction callingTool useAgentic
922.1
Elo · rank #230
Parameters
8.2B
Active params
8.2B (dense)
Context
128K tokens
Architecture
Dense transformer, GQA + SwiGLU + QK-Norm
License
Apache 2.0
Languages
119+
API price (in/out)
$0.072 / $0.287
Modalities
text
Benchmark results
Bar shows position within the tracked field; marker = field best
AIMEMath29.1%#136
best: GPT-5.2 · 100.0%
Arena-HardHuman preference79.6%#27
best: Qwen3-235B-A22B (Non-Thinking) · 96.1%
BFCL v3Agents60.2%#27
best: Hunyuan-A13B · 78.3%
C-EvalKnowledge77.9%#36
best: Qwen3.6-Plus · 93.3%
CodeforcesCoding1110#43
best: DeepSeek-V4-Pro (Think Max) · 3206
GPQA DiamondReasoning39.3%#224
best: GPT-5.6 · 94.6%
Humanity's Last ExamReasoning3.0%#137
best: Claude Sonnet 5 · 57.4%
IFBenchReasoning29.0%#67
best: MiniMax M3 · 83.0%
IFEvalReasoning83.0%#83
best: Gemma 4 26B A4B · 98.5%
LiveCodeBenchCoding22.8%#161
best: DeepSeek-V4-Pro (Think Max) · 93.5%
MATH-500Math87.4%#54
best: GPT-5 · 99.4%
MMLU-ReduxKnowledge79.5%#35
best: Qwen3.7-Max · 95.0%
MMMLUKnowledge66.9%#46
best: Gemini 3.1 Pro · 92.6%
API price $0.072/$0.287 · each benchmark row carries its own source badge (see methodology)